On Monday 01 June 2015 20:31:14 Borislav Petkov wrote:

On Mon, Jun 01, 2015 at 10:46:55AM -0700, Loc Ho wrote:

quoted

You are right... Let me just prints the error code instead. Anyone who
care will have to do post processing.

Don't forget about the usability of the driver. If a user has to go
open manuals when an error happens, you could just as well report naked
register values and have a tool decode them.

And this strategy (mcelog) turned out to be a real PITA, IMO.

quoted

Future or existent next generation may re-define them.

You could easily have a strings array of per-family or per-soc error
descriptions. For an example, take a look at drivers/edac/mce_amd.c
which decodes MCEs on all relevant AMD machines. This is much more
user-friendly than dumping register values which most people have no
idea of where to start looking (and they don't really need to).

That would require having a way to identify the SoC with a distinct
compatible string. I don't know if a name exists for X-Gene that could
be used here.

	Arnd